Hsdpa Basic

HSDPA/HSUPA High speed packet access

March 22nd 2005Salo, Finland

Harri Holma, Principal EngineerAntti Toskala, HSDPA Chief Architect

Nokia, Finland

1 © NOKIA HSDPA/HSUPA.PPT / 22-03-2005 / HH+AT

Agenda


• WCDMA R99 performance today• HSDPA standard • HSDPA performance • HSUPA standard• HSUPA performance

WCDMA Performance in Commercial Networks


WCDMA Release’99 Terminal Capability – Nokia 6630

• Data rates• WCDMA up to 384 kbps (DL) and 128 kbps (UL) • EDGE 4+2 = 236 kbps (DL) and 118 kbps (UL)

• WCDMA2100, GSM900/1800/1900• Symbian Series 60 platform• Colour display 176 x 208 pixels & 65k colours• Camera 1.3 Mpixel• Stand-by up to 11 days• Talk time up to 3 hours• Weight/volume: 127 g

Downlink 384 kbps Uplink 128 kbpsEDGE included


WCDMA Coverage – Noise Limited

-115 -110 -105 -100 -95 -90 -85 -800

50

100

150

200

250

300

350

400

CPICH RSCP [dBm]

kbps

WCDMA2100EDGE900

384 kbps

384 kbps can be provided down to approx– 105 dBm if

no other-cell interference

Similar average data rate of approx 200 kbps with

WCDMA2100 and EDGE900

3 parallel users in WCDMA


WCDMA Coverage – Interference Limited

-100 -95 -90 -85 -80 -75 -700

50

100

150

200

250

300

350

400

CPICH RSCP [dBm]

kbps

WCDMA2100EDGE900

Approx 2 x higher data rate in WCDMA than in EDGE

3 parallel users in WCDMA

Data rate other-cell interference limited at

the cell edge


WCDMA Round Trip Time (RTT)• Typical round trip time 150-200 ms for small IP packets in today’s networks • R99 round trip time expected to go below 120 ms with 10-ms TTI


End-to-end Round Trip Time (RTT) 150 – 200 ms

Radio transmission UMTS RAN + core Internet

UE RNC SGSN GGSN ServerBTS

Download of 1st WAP Page Today• 384 kbps DCH is fully used only <1 s out of 10 s• Packet setup times must get (and will get!) shorter to get full benefit from HSDPA


0.6 s

32kbps

0.9 s0.4

6 8 9 10 12

6 Radio bearer reconfiguration from 0/0 kbps to 16/32 kbps – uplink initiated

8 TCP connection establishment

9 HTTP request in uplink10 Radio bearer reconfiguration to 16/384 kbps – downlink initiated11 Page download starts over 384 kbps. TCP slow start takes about 0.7 s.

= 10 s0.7

11

12

3

0.7 s 2.0 s 1.4 s

1

RB reconfig

TCPHTTP

requestRB reconfig

Rest of the page downloaded with full 384 kbps speed, depends on the page size.

User pushes WAP home page key

RRC PDP

5

5 PDP context activation + signaling radio bearer setup + measurement control

0.4

4

4 Security mode

1

3

1.5 s

RRC connection activation

0.5

7

2 UE internal delay (in Nokia browser in Charlie)

2

384 kbps

7 UE internal delay

0.5

13

13

0.4

UE display rendering

Optimized 1st WAP Page Download• WAP page download time could be reduced from 10 s below 5 s without increasing bit rate• Setup time is reduced from 5.7 to 1.6 s, the rest is application signalling and UE delays

1st WAP Page Download Time

0.0

2.0

4.0

6.0

8.0

10.0

12.0

RAN04 Optimized

s

Display renderingDownload over 384 kbpsTCP slow startDCH upgrade DNS query and HTTP requestTCP connection establishmentUE internal delayRB reconfigurationPDP context activationSecurity modeAuthentication and cipheringRRC connection establishmentUE internal delay

10.3 s

4.7 s

1 UE Symbian delay reduced from 1.5 to 0.5 s

1

2

2 RRC setup on common channels

4

4 Direct RB allocation after PDP context

5

5 UE delay removed

6 Seamless RB upgrade6

= Setup times

3

3 Authentication done with GPRS attach already


Summary of Voice Capacity Tests• Typical voice capacity 60-70 users per cell with AMR 12.2 kbps. Up to 2x more

with lower rate AMR.

AMR 12.2 kbps users with 50% voice activity

0

10

20

30

40

50

60

70

80

Europeanop

Example lastslide

Asia, highresult

Asia, lowresult

Asia,stationary

Asia, drivetest

Voic

e us

ers

per 5

MH

z


Summary of Data Capacity Tests• Typical data capacity 600-1000 kbps per cell

0

100

200

300

400

500

600

700

800

900

1000

Packet throughputuplink

Packet throughputdownlink (64)



Agg

rega

te c

ell t

hrou

ghpu

t

Measured stationaryMeasured drive test


WCDMA High Speed Downlink PacketAccess (HSDPA) of Release 5


Reference: WCDMA for UMTS, 3rd edition, Chapter 11

Outline


• HSDPA Introduction• HSPDA Protocol Architecture• New Node B & UE functions• Modulation and coding• HSDPA & Soft Handover• HSDPA vs DCH/DSCH• HSDPA & Iub• Summary

High Speed Downlink Packet Access HSDPA

• Peak data rates increased to significantly higher than 2 Mbps; Theoretically exceeding 10 Mbps

• Packet data throughput increased 50-100% compared to 3GPP release 4

• Reduced delay from retransmissions.• Solutions

• Adaptive modulation and coding QPSK and 16-QAM• Layer 1 hybrid ARQ• Short frame 2 ms

• Schedule in 3GPP• Part of Release 5• First specifications version completed 03/02


HSDPA – General Principle


Terminal 1 (UE)

L1 Feedback

L1 Feedback

Data

Data

Downlink fast schedulingdone directly by Node B (BTS) based on knowledge of:

• UE's channel quality• UE's capability• QoS demands• Power and code resource availability• Node B buffer status

Terminal 2Users may be time and/or code multiplexed

Multi-user Diversity (Fast Scheduling)

UE2

Channel quality(CQI, Ack/Nack, TPC)


Data

Data

UE1

Multi-user selection diversity(give shared channel to “best” user)

TTI 1 TTI 2 TTI 3 TTI 4

USER 1 Es/N0USER 2 Es/N0

Scheduled user

Node-B scheduling can utilize information on the

instantaneous channel conditions for each user.


Fast Link Adaptation in HSDPA

0 20 40 60 80 100 120 140 160-202468

10121416

Time [number of TTIs]

QPSK1/4

QPSK2/4

QPSK3/4

16QAM2/4

16QAM3/4

Inst

anta

neou

s Es

No

[dB]

C/I received by UE

Link adaptation

mode

C/I varies with fading

BTS adjusts link adaptation mode with a few ms delay based on channel quality

reports from the UE


Release’99 RRM Functional Split

• Admission control on interference and power level• Initial power and SIR setting• Radio resource reservation• Air interface scheduling for common channels• DL code allocation and code three handling• Load and overload control

• Admission control on interference and power level• Initial power and SIR setting• Radio resource reservation• Air interface scheduling for common channels• DL code allocation and code three handling• Load and overload control

• Mapping RAB QoS Parameters into air interface L1/L2 parameters (incl. transport channel selection)• Air interface scheduling (for dedicated channels)• Handover Control• Outer loop power control and power balancing

• Mapping RAB QoS Parameters into air interface L1/L2 parameters (incl. transport channel selection)• Air interface scheduling (for dedicated channels)• Handover Control• Outer loop power control and power balancing

Radio network topology hidden to the CN

Radio network topology hidden to the CN

Node BServing

RNC SGSN

MSCDrift RNC

Iur IuIub

RRC

• Fast power control• Overload control

• Fast power control• Overload control


HSDPA Protocol Architecture• New MAC entity, MAC-hs added to the Node B• Layers above, such as RLC, unchanged.

WCDMA L1

UE

Iub/Iur

SRNCNode B

Uu

MACRLC

NAS

WCDMA L1

MAC-hs

TRANSPORT

FRAMEPROTOCOL

TRANSPORT

FRAMEPROTOCOL

MAC-dRLC

Iu

HSDPA user plane


Release’99 vs HSDPA RetransmissionsRel’99 DCH/DSCH Rel’5 HS-DSCH


Terminal

BTS

RNC

Packet Retransmission

RLC ACK/NACK

Retransmission

L1 ACK/NACK

Packet

HSDPA L1 Retransmissions


• The L1 retransmission procedure (Hybrid ARQ, HARQ) achieves following• L1 signaling to indicate need for retransmission -> fast round trip time facilitated

between UE and BTS• Decoder does not get rid off the received symbols when decoding fails but combines the

new transmisssion with the old one in the buffer.

• There are two ways of operating:• A) Identical retransmission (soft/chase combining): where exactly same bits are

transmitted during each transmission for the packet• B) Non-identical retransmission (incremental redundancy): Channel encoder output is

used so that 1st transmission has systematic bits and less or not parity bits and in case retransmission needed then parity bits (or more of them) form the second transmission.

New Node B functionality for HSDPA

Node BRNC Terminals

PacketsScheduler

& Buffer

ARQ &

Coding

ACK/NACK & Feedback Decoding

Flow Control

New Node B functions:• Scheduler: Terminal scheduling, Coding & Modulation selection (16QAM as

new modulation)• ARQ Retransmissions Handling• Uplink Feedback Decoding • Flow Control towards SRNC


New terminal functionality for HSDPA

RNC Node B Terminal


ARQ

Decoding

Soft Buffer

& Combining

ACK/NACK & Feedback

Generation

Packets

Flow Control

New terminal functions:• 16 QAM demodulation• ARQ Retransmissions Handling• Soft buffer & combining• Fast Uplink Feedback Generation & encoding

Adaptive Modulation – QPSK and 16QAM• Release’99 uses QPSK • HSDPA uses both QPSK and 16-QAM• 16-QAM requires also amplitude estimation from CPICH for detection

QPSK 16QAM


HSDPA - UE Categories• Theoretical peak bit rate up to 14 Mbps• 1.8 Mbps and 3.6 Mbps capability expected initially

10

9

7/8

5/6

3/4

1/2

12

11

HSDPACategory

-

-

-

3.6 Mbps

1.8 Mbps

1.2 Mbps

1.8 Mbps

0.9 Mbps

5 Codes

--36302QPSK only

--36301QPSK only

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

Modulation

14.0 Mbps

10.1 Mbps

-

-

-

-

15 Codes

-279521

-202511

7.2 Mbps144111

-72981

-72982

-72983

10 CodesTransportBlock sizeInter-TTI


HSDPA - UE Categories (cont.)• Inter-TTI interval > 1 means that terminal can not receive data in consecutive TTIs• No indication of actually being used in the market place• Other key issues are:

• 16QAM support (not part of 2 categories)• The key different in the first devices

• Size of the memory: Larger memory per class allows incremental redundancy even with full rate

• Number of HS-DSCH codes supported (5, 10 or 15)


Not permitted for the same user with Inter TTI > 1

HS-DSCH

HSDPA DL Channel Structure• High speed downlink shared channel (HS-DSCH) carries the user data in the

downlink direction, with the peak rate up to 10 Mbps • High speed shared control channel (HS-SCCH) carries the necessary physical layer

control information to enable decoding of the data on HS-DSCH• Only one HS-SCCH needed if only time multiplexing is used• DCH always running in parallel


2 ms

HS-SCCHs

HS-DSCH

……

Demodulation information

Control dataControl data

User dataUser data


HS-SCCH• The HS-SCCH is fixed data rate channel

with SF 128 (60 kbps)• The content is divided in two parts• First part carries

• Channelisation-code-set info• Modulation info

• Second Part• Transport-block size information• Hybrid-ARQ process information• Redundancy and constellation version• New data indicator

• Additionally UE identify information is used target the information to correct user

• To separate which of the 4 HS-SCCHs UE needs to decode (UE specific masking)

• Decoder matrix to be observed…..

ChannelCoding 1

HS-SCCH

Physicalchannelmapping

Ratematching 1

mux mux

Xccs Xms

Xue

X1X2

Xtbs Xhap

XrvXnd

Y

ChannelCoding 2

Ratematching 2

UEspecificmasking

Z1 Z2

S1

R1 R2

Xue

RVcoding

r s b

UE specificCRC

attachment

HS-DSCH


• The HS-DSCH uses 1 to 15 codes with fixed SF 16

• Specific parts in the channel coding chain are related to HARQ and 16QAM modulation (and some simplifications as there is no DTX, compressed mode etc…)

• Impacted by 16QAM• Interleaving (two identical interleavers

with 16QAM TTIs)• Constellation re-arrangement (same bits

not in same constellation point between retransmissions

CRC attachment

aim1,aim2,aim3,...aimA

Code block segmentation

Channel Coding

Physical channelsegmentation

PhCH#1 PhCH#P

Physical Layer Hybrid-ARQfunctionality

dim1,dim2,dim3,...dimB

oir1,oir2,oir3,...oirK

ci1,ci2,ci3,...ciE

vp,1,vp,2,vp,3,...vp,U

up,1,up,2,up,3,...up,U

w1,w2,w3,...wR

HS-DSCHInterleaving

Physical channel mapping

Constellationre-arrangement

for 16 QAM

rp,1,rp,2,rp,3,...rp,U

Bit Scrambling

bim1,bim2,bim3,...bimB

HSDPA UL Channel Structure• High Speed Dedicated Physical Control Channel (HS-DPCCH) carries the uplink

HSDPA related L1 control information to the BTS• This is parallel to the Uplink DCH• Timing from downlink packet to uplink feedback (ACK/NACK) is fixed thus

network knows for which packet the info is related to

2 ms

HS-SCCHs

HS-DSCH

7.5 slots (approx.)

HS-DPCCH

ACK/NACK Channel Quality Information

2 ms

CRC result


HS-DPCCH• The HS-DPCCH is fixed data rates channel with SF 256 • As this is BPSK channel, this gives 10 bits per slot• 1 slot used for ACK/NACK code word, 2 slots for the CQI info

• For CQI one of the 0 .. 30 values transmitted (one unused value)•(20,5) code used

i Mi,0 Mi,1 Mi,2 Mi,3 Mi,4 0 1 0 0 0 1 1 0 1 0 0 1 2 1 1 0 0 1 3 0 0 1 0 1 4 1 0 1 0 1 5 0 1 1 0 1 6 1 1 1 0 1 7 0 0 0 1 1 8 1 0 0 1 1 9 0 1 0 1 1

10 1 1 0 1 1 11 0 0 1 1 1 12 1 0 1 1 1

message to be transmitted

w0

w1

w2

w3

w4

w5

w6

w7

w8

w9

ACK 1 1 1 1 1 1 1 1 1 1

NACK 0 0 0 0 0 0 0 0 0 0

PRE 0 0 1 0 0 1 0 0 1 0

POST 0 1 0 0 1 0 0 1 0 0

Party omitted …


PAR Increase due HS-DPCCH• Terminal TX power is allowed to be reduced with low DCH uplink data rates, due

to the added parallel channel -> increased peak to average ratio when channels have close to equal power

Ratio of DPCCH/DPDCH gain factors for all values of

HS-DPCCH gain factor

Power Class 3 Power Class 4

Power(dBm)

Tol(dB)

Power(dBm)

Tol(dB)

1/15 ≤ βc/βd ≤ 12/15 +24 +1/-3 +21 +2/-2

13/15 ≤ βc/βd ≤ 15/8 +23 +2/-3 +20 +3/-2

15/7 ≤ βc/βd ≤ 15/0 +22 +3/-3 +19 +4/-2


MAC-hs Round-Trip Loop Timing• Minimum retransmission delay 12 ms


HS-SCCH

HS-PDSCH2 slots 3 slots

A = HS-DPCCH L1, MAC-hs,HS-SCCH L1

B = HS-PDSCH L1

Retransmit

Retransmit

A B

18 slots = 12 ms 2 slots

2 x Tprop + 15.5 slots

CQIA/N

Rate Matching


• Turbo encoder coding rate = 1/3.• Rate Matching is used to adapt to the

desired coding rate.• Either puncturing or repetition.• In the example, RM punctures into

rate 3/4.• Note: The systematic bits are more

important than parity bits!

Data

Turbo Encoder

SystematicParity 1Parity 2

Rate Matching (Puncturing)


Hybrid ARQ (HARQ): Chase CombiningTurbo Encoder



Original transmission Retransmission



Chase Combining (at Receiver)


Hybrid ARQ (HARQ): Incremental RedundancyTurbo Encoder




Original transmission RetransmissionSystematicParity 1Parity 2

Incremental Redundancy Combining


HSDPA Channel Quality Feedback• This will depend on the location in the cell, expected BTS TX power for HSDPA

(parameter), channel condition & receiver etc. details

High CQI reported (close to BTS, high HSDPA power)

Low CQI reported (far from BTS, low HSDPA power)


-14 -12 -10 -8 -6 -4 -2 0Tx Ec/Ior (dB)

Thr

ough

put (

kbps

)

High Throughput

Low ThroughputGeometry=0dB Geometry=5dB Geometry=10dB

Example CQI Mapping Table

• BTS can map the received CQI value for the data rate to be used in the link adaptation

• Necessary conversion to be done depending on BTS power availability

• Reference power adjustment used when quality would allow higher rate than UE capability

CQI value Transport Block Size

Number of HS-PDSCH Modulation

Reference power adjustment ∆

NIR XRV

0 N/A Out of range

1 137 1 QPSK 0

2 173 1 QPSK 0

3 233 1 QPSK 0

4 317 1 QPSK 0

5 377 1 QPSK 0

6 461 1 QPSK 0

7 650 2 QPSK 0

8 792 2 QPSK 0

9 931 2 QPSK 0

10 1262 3 QPSK 0

11 1483 3 QPSK 0

12 1742 3 QPSK 0

13 2279 4 QPSK 0

14 2583 4 QPSK 0

15 3319 5 QPSK 0

16 3565 5 16-QAM 0

17 4189 5 16-QAM 0

9600 0

(continues until 31…)38 © NOKIA HSDPA/HSUPA.PPT / 22-03-2005 / HH+AT

HSDPA & Soft Handover• In case of DCH all data is sent from all active set BTSs• In case of HSDPA, HS-DSCH sent from one BTS only, associated DCH (can be

low rate if only signaling) from all cells


Iub

RNC

Node B

Iub

Node B

Node B

RNC

DCH +HS-DSCH

DCH

DCH

DCH

Node B

HSDPA & Soft Handover (cont.)• The intra-frequency measurement event ID is modified• Now a measurement report will be initiated when the best serving cell changed

(parameters to have some hysteresis)• This is needed to initiate the HS-DSCH serving cell change even when active set is

unchanged• In case serving cell change, RLC layer (in the RNC) will handle unfinished ARQ

processes when Node B memory is flushed.


IubRNC

Node B

DCH +HS-DSCH

DCH

Node B

HS-DSCH vs. DSCH• DSCH of Release’99 uses variable spreading factor and fast power control

(otherwise like DCH)• Principle agreement also to remove DSCH from Release 5 and onwards

from 3GPP specifications as part of feature clean-up (CRs for 06/05 specs)• HS-DSCH of Release 5 uses adaptive modulation and coding and L1 Hybrid

ARQ

Feature

Variable spreading factor

Fast power control

Adaptive modulation and coding

Fast L1 HARQ

DSCH

Yes

Yes

No

No

HS-DSCH

No

No

Yes

Yes


HSDPA & DCH Resource Sharing


Common channels

DCH RT

DCH NRT

HSDPA NRT

PtxTarget

Max power

Power control head-room

Non-controllable power

Controllable power

Total transmittedcarrier power

NEW non-HSDPApower measurements

Power measurementsfrom the Node-B to

the RNC

Node-B Tx power

In addition to power also code resource

shared!

HSDPA vs DCH Code Space Usage• With Downlink DCH Time multiplexed DPCCH and DPDCH• Variable rate with DTX -> Code space not released due inactivity• Problematic for high bit rates -> Current highest DL rate 384 kbps

FULL RATE

Data TPC


PilotData

Slot 0.667 ms = 2/3 ms

TPC

DPDCH DPDCH

Data

DPCCH DPCCH

TFCI

PilotDataTFCI

Slot 0.667 ms = 2/3 msDPDCH DPCCH DPDCH DPCCH

HALF RATE

DTX

HSDPA vs DCH code space usage (cont.)• With HSDPA code resource (except what is needed for DCH) shared with 2 ms

resolution -> Optimum code resource use• Also no soft handover related extra code use


C0(0) = [ 1 ]

C1(0) = [ 1 1 ]

C1(1) = [ 1 0 ]

C2(0) = [ 1 1 1 1 ]

C2(1) = [ 1 1 0 0 ]

C2(2) = [ 1 0 1 0 ]

C2(3) = [ 1 0 0 1 ]

C3(0) = [ 1 1 1 1 1 1 1 1 ]

C3(1) = [ 1 1 1 1 0 0 0 0 ]

. . .

. . .

Spreading factor:

SF = 1 SF = 2 SF = 4 SF = 8

C3(2) = [ 1 1 0 0 1 1 0 0 ]

C3(3) = [ 1 1 0 0 0 0 1 1]

. . .

. . .

C3(4) = [ 1 0 1 0 1 0 1 0 ]

C3(5) = [ 1 0 1 0 0 1 0 1 ]

. . .

. . .

C3(6) = [ 1 0 0 1 1 0 0 1 ]

C3(7) = [ 1 0 0 1 0 1 1 0 ]

. . .

. . .

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

= Allocated code

= Code which cannotbe allocated at the sametime as C3(1)

= Code which canbe allocated at the sametime as C3(1)

These codes cannot be used at the same

time as C3(1)

HSDPA & Iub• HSDPA improves Iub efficiency compared to Release’99 packet data since HSDPA is

a time shared channel with a flow control in Iub• Release’99 requires dedicated resources from RNC to UE. Those resources are not

fully utilized during TCP slow start, during data rate variations or during inactivity timer

• Additionally, HSDPA does not use soft handover ⇒ no need for soft handover overhead in Iub


= User 1= User 2= User 3

Iub link 1

Iub link 2

1 2 HSDPA Iubcapacity

Iub efficiently utilized by HSDPA

1 2

= TCP slow start1= Inactivity timer2

HSDPA & Iub (Cont.)• The needed Iub capacity is NOT

equal to air interface peak rate (e.g. 1.8 or 7.2 Mbps)

• The momentary data rate over 2 ms can be higher than the Iub capacity available, as BTS has buffers and is scheduling to multiple users

• Iub loading thus more related to the actual air interface capacity

0.01 0.1 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

180% power and 15 codes allocated to HSDPA service

Instantaneous (per 2 ms) user throughput [Mbps]

Cum

ulat

ive

distr

ibut

ion

func

tion

[-]

Macrocell/Veh A/3kmph

Microcell/Ped A/3kmph

Average data rate avaible


HSDPA - Summary


• Multi-code operation combined with lower coding rates and fast HARQ improves link performance at cell edge (low SIR)

• Multi-code operation combined with increased coding rates (e.g. 3/4) fully utilize favorable radio environments (high SIR) without running into code shortage.

• HSDPA is backwards compatible and can be introduced gradually in the network.

• Retransmission and scheduling into Node B• -> reduces (re-)transmission delays; Improves QoS control.

Freely configurable transmissionHSDPA is a natural capacity evolution to WCDMAand an enabler for higher speed data services

HSDPA Release 6 improvements


• The following items are added in Release 6 specifications• Fractional DPCH

• To avoid need for full DCH in downlink when only PS data (SRB in HS-DSCH), shared DL channel to provide uplink TPC commands

• Pre/Post scheme• To avoid DTX detection in BTS for ACK/NACK

commands -> smaller power offsets for the ACK/NACK –> better uplink range

• RX diversity and improved receiver performance• Latter still on-going

• L2/L3 Signaling improvements on-going as well…. • Proposed e.g. combined active set update and HS-DSCH

serving cell change

• Mandatory for a HSDPA capable Release 6 UE, RX diversity and improved receiver optional from the spec …

HSDPA Performance


HSDPA RLC Layer Data Rates• Max RLC layer throughput shown below• Max application layer throughput can be very close to RLC throughput

• Difference <5% mainly due to IP headers

5 codes QPSK 10 x 320 3440 1.6 Mbps

# of codes Modulation RLC blocks per 2 ms TTI1

Transport block1

Max RLC data rate

5 codes 16-QAM 21 x 320 7168 3.36 Mbps

10 codes 16-QAM 42 x 320 14155 6.72 Mbps

15 codes 16-QAM 60 x 320 20251 9.6 Mbps

15 codes 16-QAM 83 x 320 27952 13.3 Mbps

12

UE category

5/6

7/8

9

10

1RLC block size of 320 bits assumed2Includes user plane + RLC headers + MAC headers + padding


Nokia HSDPA RAN Measurements


Nokia HSDPA RAN Throughput [kbps]• Application throughput matches with the theoretical calculation • >1.5 Mbps measured with FTP download

Throughput [kbps]

0200400600800

100012001400160018002000

L1 throughput(theory)

RLC throughput(theory)

Applicationthroughput

(theory)

Applicationthroughput(measured)

1.8 Mbps5-code QPSK

1.6 Mbps 1.53 Mbps


Nokia HSDPA RAN Latency [ms]• Measured round trip time 94 ms with 20-ms uplink TTI• Approx 20 ms faster expected with 10-ms uplink TTI


020406080

100120140160180200

RAN1.5ED2 RAN04 (20-ms UL+DL)

RAN04 (20-ms UL 10-ms

DL)

RAN04 (10-ms UL+DL)

3GPP R5HSDPA (20-ms UL 2-ms

DL)

3GPP R5HSDPA (10-ms UL 2-ms

DL)

3GPP R6HSUPA (2-ms UL+DL)

ms

CalculatedMeasured

94 ms measured with 20-ms uplink

Measured with Ubinetics

System Control Algorithms


Link Adaptation and Power Control


• HS-DSCH link adaptation has typically two loops• A) Inner loop based on CQI reports from UE• B) Outer loop in Node-B to control the BLER of the inner loop

• HS-SCCH power control has typically two loops as well• A) Inner loop based on the fast power control commands from UE• B) Outer loop in Node-B to control the BLER of the inner loop

• These algorithms are not standardized

HS-DSCH link adaptation

HS-SCCH power control

Link Adaptation


A Two-Step HS-DSCH Link Adaptation (1)


• Inner loop algorithms• LA estimates based on UE feedback information, i.e. CQI reports.• LA estimates based on Node-B measurements = UE power control commands• Hybrid schemes using both UE/Node-B measurements.

• Outer loop algorithms• The goal is to compensate for any bias introduced by the inner loop algorithm.• Bias might be introduced due to offsets in relative UE performance caused by improved

receiver architecture, etc.• The outer loop is based on Ack/Nack.• The outer loop controls the residual BLER after 1st retransmissions (=2nd transmission) • The 1st retransmission is used to take into account the gain from HARQ.

A Two-Step HS-DSCH Link Adaptation (2)

RNC Node-B

HS-DSCH

CQI PDCHUL HS-DPCCHdetection

Inner loopLA algorithm

Outer loopLA algorithm

Handovercontrol

CQIsettings


CQI Based Inner Loop Procedure


• Step 1: CQI offset compensation• Compensate for HS-PDSCH transmit power mismatch, if the CQI report at the UE is

derived for a different power level than the one used at the Node-B.

• Step 2: Mapping of CQI report to feasible LA estimate• The offset compensated CQI report may not correspond to a feasible LA estimate, if for

instance the number of suggested multi-codes exceeds the number of HS-PDSCH multi-codes available at the Node-B.

• Step 3: Transport block size adjustment• Adjust the transport block size for the feasible LA estimate according to the number of

available bits to be transmitted to the UE, i.e., perform a simple “rounding” operation.

Outerloop Algorithm Based on Ack/Nack's


Start

Set A equal to the predefined initial value

Ackreceived from re-transmission of

block ?

first

Ackreceived from

transmission ofblock ?

first

Decrease Aby "AStepDown" [dB]

Increase Aby "AStepUp" [dB]

Yes

Yes

No

No

Increase set-point if:

Nack on 2nd trans.

Decrease set-point if:

Ack on 1st or 2nd trans.

Ratio between up/down step determines the residual BLER, I.e. similar as conventional outerloop power control.

HS-SCCH Power Control


HS-SCCH PC Summary

RNC Node-B

HS-SCCH transmit power

CQI PDCHUL HS-DPCCHdetection

SHO: On\Off

HS-SCCH power offset Inner loop

PC algorithmOuter loop

PC algorithm

Handovercontrol

HSDPAoffset

P /0 Por

1

β

Inner loop based on CQI


Outer Loop PC for the HS-SCCH


• The outer loop for HS-SCCH is similar to the standard outer loop PC algorithm for dedicated channels.

• Uplink detection of DTX on the HS-DPCCH (when expecting an Ack/Nack) is assumed to indicate that the HS-SCCH has been erroneously decoded.

• Algorithm• DTX received : Increase target with Pup decibels• Otherwise : Decrease target with Pdown decibels

• The outer loop algorithm is UE specific• Given this approach, the HS-SCCH average BLER converges to

1

1

+=

down

up

PPBLER

System Performance and Simulations


HSDPA Performance

Link performanceLink performance How far from Shannon limit?


Macro cell performance – single

user per cell

Macro cell performance – single

user per cellWhat can be achieved in macro cells if all resources given to single user?

Scheduler performance –

multiple users per cell

Scheduler performance –

multiple users per cellHow cell resources can be shared between simultaneous users?

Application performance – end

user view

Application performance – end

user viewWhat is e2e application performance over HSDPA?

HSDPA Link Performance with Turbo Coding Approaches Shannon Limit

Higher bit rates can be obtained only with more antennas (MIMO) and/or wider bandwidth.

Sup

porte

d ef

fect

ive

data

rate

[Mbp

s]

0.1

1.0

10.0

-15 -10 -5 0 5 10 15 20

16QAM

0.01

Instantaneous HS-DSCH C/I before processing gain [dB]

QPSK

HSDPA linkadaptation curve

Shannon limit:3.84MHz*log (1+C/I)2

15 HS-PDSCH allocation(Rake, Pedestrian-A, 3km/h)


HSDPA Bit Rates

0.01 0.1 1 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

180% power and 15 codes allocated to HSDPA service

Instantaneous (per 2 ms) user throughput [Mbps]

Cum

ulat

ive

distr

ibut

ion

func

tion

[-]

Macrocell/Veh A/3kmph

Microcell/Ped A/3kmph

Mean bit rate1.5 Mbps in macro cells

Mean bit rate>5 Mbps in micro cells


Bit Rate Coverage – Interference Limited Case

-100 -98 -96 -94 -92 -90 -88 -86 -84 -82 -800

200

400

600

800

1000

1200

1400

1600

1800

2000

CPICH RSCP [dBm]

kbps

HSDPA2100WCDMA2100EDGE900

3 parallel users in WCDMA and in HSDPA

EDGE900

Average data ratesHSDPA2100 1920 kbps (1 user)HSDPA2100 640 kbps (3 users)WCDMA2100 250 kbps (3 users)EDGE900 140 kbps

Average data ratesHSDPA2100 1920 kbps (1 user)HSDPA2100 640 kbps (3 users)WCDMA2100 250 kbps (3 users)EDGE900 140 kbps

HSDPA2100 (3 users)

WCDMA2100 (3 users)


HSDPA Coverage – Interference Limited


0

1000

2000

3000

4000

5000

6000

7000

8000 Single user assumed on HS-DSCH

7.2 Mbps 2-rx equalizer other

cells fully loaded

3.6 Mbps 1-rx Rake with other cells fully loaded

Average data rate 1-3 Mbps

kbps

Base station Cell edge

7.2 Mbps 2-rx equalizer other cells low loaded

Advanced terminals push

data rates higher

Average user data rate

0

500

1000

1500

2000

2500

3000

2 4 6 8 10 12

HS-DSCH power [W] out of 16-W BTS power

kbps

10-code, 2-eq

10-code, 1-eq

5-code Rake

Average HSDPA Bit Rates with Shared R99 Carrier

0.5-1.5 Mbps in macro cells if carrier

shared with R99

Single user assumed on HS-DSCH


Packet Scheduling Strategies

Fair Throughput

Fair timeC/I based

Operator adjustable slopes for different service and user classes

Min

. tar

get T

hrou

ghpu

t

Tradeoff between cellthroughput, coverage

and user fairnessSpectral Efficiency


Throughput coverage of C/I based scheduling

Throughput coverage of fair throughput scheduling

Throughput

high

low

Low coverage of high data rates

Packet Scheduling Strategies

PS Method

Fair throughput(FT)

*) Note that actual resource allocation will be heavily influenced by QoS/delay requirements and user prioritization schemes!

Fair resource(FR)

C/I based(CI)

Serve order Allocation method

Round robin in random order

Resources according to samethroughput (up to max. alloc. time)

Round robin in random order

Same resources (time/power/codes) per allocation time

Served according to highest slow-averaged channel

quality


Scheduling rate

Slow(avg. over 20-100ms)



*) Note that actual resource allocation will be heavily influenced by QoS/delay requirements and user prioritization schemes!

Max C/I based(M-CI)

Served according to highest instantaneous channel

quality


Fast(Per-TTI basis)

Proportional Fair throughput (P-FT)

Proportional Fair resource (P-FR)

Served according to highest relative instantaneous channel quality (RICQ)

Resources according to samethroughput (up to max. alloc. time)

Served according to highest relative instantaneous channel quality (RICQ)


Fast(Per-TTI basis)

Fast(Per-TTI basis)


Multi-user Diversity (Fast Scheduling)

UE2



Data

Data

UE1

Multi-user selection diversity(give shared channel to “best” user)

TTI 1 TTI 2 TTI 3 TTI 4

USER 1 Es/N0USER 2 Es/N0

Scheduled user

Node-B scheduling can utilize information on the

instantaneous channel conditions for each user.


Proportional Fair Algorithm• Principle is to schedule the user who currently has the highest ratio of instantaneous

throughput to average throughput. The averaging time is typically a few 100 ms.• The user with the highest selection metric at a given time is selected for scheduling

in the following TTI


[ ][ ]nTnRM

k

kk ≡ Rk = instantaneous supported data rate for user k

Tk = average throughput for user k

Proportional Fair Scheduling as a Function of Number of Users


Avg.

cel

l cap

acity

gai

nof

PF

PS

ove

r FR

PS

[dB

]

User diversity order, UDO [-]0 5 10 15 20 25 30

0.8

1

1.2

1.4

1.6

1.8

2

2.2

2.4

Macrocell/Ped-A (3km/h), PCDE=-36dBMacrocell/Veh-A (30km/h), PCDE=-36dBMicrocell/Ped-A (3km/h), PCDE=-36dB

• Significant gain from proportional fair scheduling if at least 3-5 simultaneous users.

HSDPA Cell ThroughputDynamic System Simulation Macro Cells Vehicular A Channel

0

500

1000

1500

2000

2500

3000

3500

4000

4500

Rake 1-ant Equalizer 1-ant Rake 2-ant Equalizer 2-ant

kbps

Round robin 5 codesRound robin 10 codesProportional fair 5 codesProportional fair 10 codesProportional fair 15 codes

1 Mbps

2 Mbps

4 Mbps

• 5-code BTS and single antenna UE Rake provides 1 Mbps• 10-code BTS and single antenna UE provides 2 Mbps• 15-code BTS and dual antenna UE provides 4 Mbps


Transport Block Size Selection Statistics

Break point for16QAM selection

10 HS-PDSCH codes 16-QAM selection probability <5%

in macro cells

16-QAM selection probability 25%

in micro cells


Co-Existence of R99 and HSDPA


Co-existence of R99 and HSDPA= R5 HSDPA • HSDPA can be introduced to the network with shared or

with dedicated frequency = R99 DCH

f1f1• Carrier shared between HSDPA and R99• Operator definable or dynamic resource sharing between

HSDPA and R99


f1f1

f2f2 • Dedicated HSDPA carrier• HSDPA UE directed to HSDPA carrier

f1f1

f2f2 • HSDPA carrier, which can also be used for R99 traffic in case of R99 high load

• UE moved by RRC connection setup or by handovers

HSDPA and Iub Capacity


HSDPA Improves Iub Efficiency by 50-100%• In total, HSDPA improves Iub efficiency 50-100% compared to Release’99• HSDPA requires more E1 lines? Yes, if we want to show higher peak rates. 2 x E1

is required to provide peak rate of 2 Mbps and 3 x E1 for 3.6 Mbps.• HSDPA can be operated with 1 x E1? Yes, and it even improves Iub efficiency but

the peak data rate will be approx 1 Mbps. Also, the operator may want to provide a minimum HSDPA throughput and single E1 is not enough in that case.


InactivityInactivity

Soft HOSoft HOHSDPA

user dataHSDPA

user dataRelease’99user data

Release’99user data

Common ch

DCH HSDPA

Common ch Common chCommon ch

E1 capacityIub capacity requirement

HSDPA RF vs Iub efficiency

60%

65%

70%

75%

80%

85%

90%

95%

100%

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8Iub capacity for 1+1+1 site [Mbps]

RF utilization [%]Iub utilization [%]

Trade-off Between HSDPA RF and IubCapacity

High Iub efficiency (95%) requires some overhead (15%) in air

interface dimensioning

High RF efficiency (95%) requires some

overhead (25%) in Iub dimensioning


End User Capacity and Cost


HSDPA Enables Mass Market >1 GB/Sub/Month

0.0

0.5

1.0

1.5

2.0

2.5

300 400 500 600 700 800 900 1000Subscribers per site

GB

/sub

/mon

th

HSDPA 2-rx UE

HSDPA 1-rx UE

HSDPA with 1rx UE 2.0 Mbps/cellHSDPA with 2rx UE 3.6 Mbps/cellBusy hour share 20%Busy hour utilization 80%Spectrum 2 x 5 MHz = 1 carrier dedicated for HSDPA

1 GB/month for 300-600 subs

1-carrier HSDPA assumed


Cost of Data Delivery (Network Capex)

Delivery cost of downloaded GB

0.00.51.01.52.02.53.03.54.04.55.0

14 16 18 20 22 24 26 28 30 32 34 36 38 40

Price per TRX [k€]

[EU

RO

€/G

B]

<4 €/GB data

Depreciation over 6 yearsOther assumptions from previous slide


VoIP over HSPA


VoIP Usage Growing Fast in Fixed Internet

• Close to 100 Mdownloads and counting...

• Also Pocket PCversion, ca. 2 MB


VoIP Drivers and Challenges in Cellular


• Drivers1. Enterprise VoIP extension to wide area2. Rich call interworking3. Plain vanilla VoIP for network simplicity

• Challenges1. End-to-end delay capability2. Bandwidth and delay during high load3. Efficiency in the air interface and Iub

RNC SGSN/GGSNBTS

Potential VoIP Capacity on HSDPA• HSDPA VoIP can increase voice capacity – but at the expense of end-to-end delay• 80-150 ms was allowed for queuing and transmission in the results below, which

causes >250 ms worst case mobile-to-mobile delay• At maximum load HSDPA VoIP delay was about 100 ms more than in CS voice.

During low load e2e delay with HSDPA VoIP is slightly less than with CS voiceAMR 12.2 kbps voice users per cell

0

20

40

60

80

100

120

R99 CS voice HSDPA VoIP withround robin

HSDPA VoIP withproportional fair

Voi

ce u

sers

150 ms delay80 ms delay

Queuing + transmission

delay in downlink


Radio Technology Evolution


Round trip time [ms]400 300 200

End user bit rate [kbps]

GPRS500-700ms40kbps

EGPRS, Rel. 4 100-200kbps200-500ms

EGPRS, Rel. 99 100-200kbps500-700ms

HSPA1-2 Mbps50-100ms

WCDMA, R99 200-300kbps150-200ms

EVDO Rev 0400 kbps200-300 ms

ADSL1-2 Mbps<50ms

VoIP feasible from e2e performance point of view

3Mbps

1Mbps

300 kbps

100 kbps

30 kbps

100 0600 500

Conclusions on HSDPA Performance


• End user capabilities• Peak data rate 3.6 – 14.4 Mbps• Macro cell data rates 1.0-2.5 Mbps with dedicated carrier and 0.5-1.5 Mbps when shared

carrier with R99• Latency <100 ms• Simultaneous CS voice + PS data on the same connection

• Capacity and cost• Cell throughput 2 Mbps/carrier • User capacity 1 GB/month/subscriber for 300-600 subs/site• Delivery cost <4 €/GB for network capex

• Integration with WCDMA• HSDPA can co-exist on the same carrier as WCDMA• WCDMA integrates efficient voice network to the low latency high bit rate HSDPA

network

Mobile Broadband Access Landscape

Mobile broadband wireless access

CellularFixed broadband wireless 802.16

Wireless DSLFlat rate

Mobile premiumValue added services

EDGE, WCDMA, cdma

EV-DOHSDPA

802.16e

LAN

802.11 WiFi

Proprietary extensionsFlarion

Proprietary



Technology Positioning - Use Cases1 Fixed BWA “Wireless DSL”

• User uses wireless broadband service from fixed location, home or office, in similar way as DSL or cable modem access

• Usually requires non-movable, outdoor antenna for coverage

2 Portable BWA “Laptop/PDA access from park”• User uses wireless broadband service from any location within coverage

area• Mobility of the user between cells is not supported (no handovers)• Use case is very similar to that of hot spot WLAN access

3 Mobile BWA “Smart phone access from car”

• User uses wireless broadband service from any location within coverage area including the mobility of the user (e.g. up to 250 km/h)

• Use case is very similar to that of 3G cellular data

Spectrum

..

850

1800

19002.1 GHz

2.6 GHz3.4 GHz5 GHz

2x60 MHz

New 3G band in Europe by 2008190 MHz

Unlicenced

Mainstream WCDMA

1700 2x45 MHz New USA 3G band by 2006

Fixed wirelss access band in Europe

= WCDMA/HSPA band with products today= WCDMA/HSPA band - future

2x75 MHz

2x60 MHz

900 2x35 MHz

= Expected bands for 802.16

Europe and Asia

2.5 GHz MMDS in USA190 MHz

US PCS band

Americas, Japan, Asia2x25 MHz

Europe and Asia

Smallest coverage

Best coverage


HSDPA and EV-DO


• HSDPA and EV-DO have similar technical solutions, which is leads to similar spectral efficiency for best effort data

• Fast retransmissions• Hybrid ARQ• Adaptive modulation and coding

• HSDPA allows simultaneous voice and packet data• Simultaneous CS voice + PS video possible in HSDPA• EV-DO Rev.A must use VoIP to carry voice

• HSDPA has higher bandwidth, which brings high user bit rates• EV-DO downlink typical bit rate during low loading 300-500 kbps• HSDPA downlink typical bit rate during low loading 1-2 Mbps

WCDMA High Speed Uplink Packet Access (HSUPA) of Release 6


Reference: WCDMA for UMTS, 3rd edition, Chapter 11

Outline


• HSUPA Introduction• HSUDA Protocol Architecture• HSUPA Retransmissions• HSUPA Peak Bit Rates• HSUPA UE Capabilites• HSUPA Channel Stuctures Uplink and Downlink• Summary

High Speed Uplink Packet Access HSUPA

• Peak data rates increased to significantly higher than 2 Mbps; Theoretically reaching 5.8 Mbps

• Packet data throughput increased, though not quite high numbers expected as with HSDPA

• Reduced delay from retransmissions.• Solutions

• Layer 1 hybrid ARQ• Node B based scheduling for uplink• Frame sizes 2ms & 10 ms

• Schedule in 3GPP• Part of Release 6• First specifications version completed 12/04• In 3GPP specs with the name Enhanced uplink DCH (E-DCH)


HSUPA Status in 3GPP• Study item for 3GPP Release 6 completed 03/04• First versions of HSUPA specs published 12/04• Work item completion date 03/05• Remaining open issues closed 06-09/05


2003 2004 2005 2006 2007

3GPP study item

Study item completed

1st version in 3GPP spec

Official work item completion date

HSUPA Protocol Architecture• New MAC entity, MAC-e added to the Node B• In RNC MAC-es handling packet in-sequence delivery & soft handover combining• Layers above, such as RLC, unchanged -> this required MAC-es to perform

reordering for packets

RLC

WCDMA L1

UE

Iub/Iur

SRNCNode B

Uu

MAC

NAS

HSUPA user plane

WCDMA L1

MAC-e

TRANSPORT

FRAMEPROTOCOL

TRANSPORT

FRAMEPROTOCOL

MAC-esMAC-d

IuRLC

MAC-es/e

RLC


Node B Controlled HSUPA Scheduling

Data packet

+ possible retransmissions

+ control for scheduling

ACK/NAK

+ control

Node B RNCUE

Mac-es

Iub

Mac-e

New Node B functions:Uplink packet data schedulingHARQ control: ACK/NAKs

New Iub signalling

New L1 signalling

• Target is to shorten the packet scheduling period ⇒ packet scheduler is able to track burstiness of source application


Fast Scheduling Reduces Noise Variance

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

NR [dB]

PDF of the NR per BTS

Mean: 4.1 (dB) StD: 1.2 (dB)

Mean: 5 (dB) StD: 0.72 (dB)

RNC sheduling Node B scheduling

• Faster scheduling reduces noise rise variations

⇒ Less headroom needed⇒ Cell capacity and user data

rates are increased• With low loaded uplink, the

users may get significantly higher data rates as much more aggressive data rates can be granted to UEs

Operation point can be increased because variance is reduced.


DCH vs E-DCH RetransmissionsDCH E-DCH


RNC

Terminal

BTS

Packet Retransmission

RLC ACK/NACK

Combining of packet and

retransmission

Packet (1st TX) Retransmission

L1 ACK/NACK

Fast Hybrid-ARQ between UE and BTS• Fast ACK/NAK from BTS• N-process Stop-And-Wait (SAW) HARQ (similar to HSDPA)• short round trip delay => lower total delay• Chase combining or Incremental Redundancy, soft buffering in BTS• In SHO, each BTS sends ACK, retransmission if no ACKs


UEBTSRNC Correctly received packet

HARQ control and soft combining

ACK/NAK

ACK/NAK

E-DCH data

E-DCH data


Feedback from UE to BTS/RNC• For the RNC UE provides the following information (MAC layer)

• UE buffer occupancy: how much data buffers • Available transmission power resource

• For the BTS, the following is provided on E-DPCCH (L1)• Information of the data rate for the frame (E-TFI)• Information of the redundancy version for the packet

• Timing is known thus BTS knows which ARQ channel to expect

• Happy bit: Is the current data rate satisfactory• UE would not be happy of the data rate if it could transmit

with higher rate due:– Transmission power left– Data left in the buffers (for more than X TTIs, X as

parameter)

BTS

RNC

E-DPCCH

MAC-PDU

Adaptive Modulation – Why not with HSUPA?


• In the downlink direction, BTS has limited power control dynamics, in the order of 10-20 dB

• However in the uplink received power level is kept constant with fast closed loop power control with more than 70 dB dynamic range

• Thus there are no times when there would be “free lunch” to use higher order modulation

• Other reasons why higher order modulation sounds interesting*…• A) To get more bits per given bandwidth - > range problem• B) To reduce the terminal peak to average ratio as having multicodes in use later ->

This would have resulted to higher average power even if the PAR would have been smallel -> less capacity

* When searching from the web HSUPA info, often one sees adaptive modulationas part of the story. This is based on the outdated stuff from the study item phase

HSUPA - UE Categories• HSUPA uses BPSK modulation with multicode transmission to achieve high data

rates, 6 different UE categories defined, key element number of codes supported and whether 2 ms TTI is supported or not

• Theoretical peak bit rate up to 5.76 Mbps• 1.46 Mbps capability expected initially

11520

20000

20000

5837

20000

14592

2919

14592

7296

Transport Block size

2 Mbps102 x SF24

2.9 Mbps22 x SF24

1.46 Mbps102 x SF42

1.46 Mbps22 x SF42

2 Mbps102xSF2 + 2xSF46

6

5

3

1

HSUPACategory

2

10

10

10

TTI

2xSF2 + 2xSF4

2 x SF2

2 x SF4

1 x SF4

Codes x Spreading

5.76 Mbps

2 Mbps

1.46 Mbps

0.73 Mbps

Data rate


HSUPA Channel Structure - Uplink• E-DPDCH, (E-DCH Dedicated Physical

Data Channel) - carries the user data in the uplink

direction, with the peak rate reaching up to 5.76 Mbps

- codes are not shared with HSUPA (no code shortage in the uplink)

• E-DPCCH (E-DCH Dedicated Physical Control Channel )

- Parallel physical layer control channel - carrying e.g. ARQ information needed

for the BTS decoding process + UE txbuffer status

- Normal DCH operated in parallel (DPCCH always and DPDCH if there is data)

Σ I+jQ

Se-dpch

ced,1 βed,1

E-DPDCH1

iqed,1

ced,k βed,k

E-DPDCHk

iqed,k

ced,K βed,K

E-DPDCHK

iqed,K

cec βec

E-DPCCH

iqec

.

.

.

.

.

.

.

.


HSUPA Channel Structure – Uplink (cont)


• With E-DPDCH, there are two transmission time intervals (TTIs):a) 10 ms TTI (DCH has 10, 20, 40 and 80 ms)

- Best for range point of view, typically better capacity than 2 ms b) 2 ms TTI

- Suffers in range, thus with larger cells not usable close to cell edge- Can be used to delay reduction

• E-DPDCH uses multicode transmission • Up to 4 parallel codes (see next slide)

DataNdata bits

E -DPDCH

1 radio frame: T = 10 ms

Slot #0 Slot #1 Slot #i Slot #14Slot #2

1 subframe = 2 ms

Tslot

= 2560 chips, Ndata

= 1 0*2 k+2 bits (k=0...5)

HSUPA Channel Structure – Uplink (cont)• The smallest data rate with single SF 4 code

• Then two SF 4 codes• Then two SF 2 codes• And highest data rate with 2 times SF 2 & 2 times SF 4 (Highest PAR)

• Compare to Release’99 where after reaching SF 4 (e.g. 384 kbps), higher data rates would be by adding more SF 4 codes (up to 6)

Slot Format #i

Channel Bit Rate (kbps)

SF Bits/ Frame

Bits/ Subframe

Bits/SlotNdata

0 60 64 600 120 40

1 120 32 1200 240 80

2 240 16 2400 480 160

3 480 8 4800 960 320

4 960 4 9600 1920 640

5 1920 2 19200 3840 1280

Slot Format #i


SF Bits/ Frame

Bits/ Subframe

Bits/SlotNdata

0 60 64 600 120 40

1 120 32 1200 240 80

2 240 16 2400 480 160

3 480 8 4800 960 320

4 960 4 9600 1920 640

5 1920 2 19200 3840 1280

Slot Format #i

Slot Format #i



SFSF Bits/ FrameBits/

FrameBits/

SubframeBits/

SubframeBits/Slot

Ndata

Bits/SlotNdata

00 6060 6464 600600 120120 4040

11 120120 3232 12001200 240240 8080

22 240240 1616 24002400 480480 160160

33 480480 88 48004800 960960 320320

44 960960 44 96009600 19201920 640640

55 19201920 22 1920019200 38403840 12801280

E-DPDCH slot formats


E-DCH Coding


• One transport block once per TTI• CRC attachment: 24 bit CRC • Code block segmentation similar to the rel’99/4/5

DCHs, the value of maximum code block size is 5114 for turbo coding

• Channel coding: rate 1/3 turbo coding (as rel’99)• Physical layer HARQ functionality and rate

matching: SF and number of needed E-DPDCHsare determined (similar to rel’99/4/5, but different parameters) and the physical layer HARQ functionality (see next slide)

• Physical Channel Segmentation: similar to the rel’99/4/5 DCHs, when more than one E-DPDCH is used, physical channel segmentation distributes the bits among the different physical channels

• Interleaving and physical channel mapping: similar to the rel’99/4/5 DCHs, rate matching output bits are interleaved and mapped to the allocated E-DPDCH(s)

Transport channel processing for E-DCH

CRC attachment

aim1,aim2,aim3,...,aimA

Code block segmentation

Channel Coding

Physical ChannelSegmentation

Physical channel(s)

Physical Layer Hybrid-ARQ

functionality/Rate matching

oir1,oir2,oir3,...,oirK

ci1,ci2,ci3,...,ciE

up,1,up,2,up,3,...,up,U(p)

s 1,s 2,s 3,...,s R

bim1,bim2,bim3,...,bimB

Interleaving &


E-DCH Channel Coding (cont.)

• Bit separation (similar to rel’99/4/5 turbo coded DCHs)• Rate matching (similar to rel’99/4/5 DCHs)

• PLmax is equal to 0.44 for all E-DCH UE categories except the highest E-DCH UE category, for which PLmax is equal to 0.33

• Parameters depend on the redundancy version (RV) parameters s and r similar to HS-DSCH HARQ second rate matching stage

• Higher layers (RNC) configure, if only RV=0 is used or if RV can be 0, 1, 2 or 3• Bit collection, similar to HS-DSCH HARQ

Systematicbits

Parity 1bits

Parity2bits

RM_P1_2

RM_P2_2

RM_S

Rate Matching

Nsys

Np1

Np2

Nt,sys

Nt,p1

Nt,p2

bitseparation

Ne,j bit

collection

Ne,data,j

E-DCH hybrid ARQ functionality112 © NOKIA HSDPA/HSUPA.PPT / 22-03-2005 / HH+AT

E-DPCCH Coding

E-DPCCH

Physicalchannelmapping

Multiplexing

xh,1 xrsn,1, xrsn,2

ChannelCoding

xtfci,1, xtfci,2,..., xtfci,7

x1, x2,..., x10

z0, z1,..., z29


• Control information on E-DPCCH is multiplexed: •E-TFCI information (7 bits)•Retransmission sequence number (RSN, 2 bits)•‘Happy bit’ (Rate Request, 1 bit)

• Channel Coding: a sub-code of the second order Reed-Muller Code (similar to rel’99/4/5 TFCI coding)

• Physical Channel Mapping: similar to rel’99/4/5, channel coding output bits are mapped to the allocated E-DPCCH

Coding chain for E-DPCCH

HSUPA Channel Structure - Downlink


• The following downlink control channels are defined• E-RGCH (E-DCH Relative Grant Channel)

• Relative grants, to set data rate up/down or hold constant• E-AGCH (E-DCH Absolute Grant Channel)

• Absolute grants, not only up/down• E-HICH (E-DCH HARQ Acknowledgement Indicator Channel)

• Information on the packet transmission station (ACK/NACK)• Shares the same code with E-RGCH

• E-RGCH and E-HICH share the same SF 128 code• For one UE only one is relevant but single code can serve serveral UEs

Slot #14

Tslot = 2560 chip

bi,39bi,1bi,0

Slot #0 Slot #1 Slot #2 Slot #i

1 radio frame, T f = 10 ms

1 subframe = 2 ms

E-HICH (E-DCH Hybrid ARQ Indicator Channel)

• E-DCH hybrid ARQ acknowledgement indicator a is mapped on E-HICH:• ACK: +1• NACK: –1 in cells belonging to the same Radio Link Set (RLS) as the E-DCH

serving cell and 0 in other cells (DTX)

• When Node B has not detected E-DPCCH in uplink, no ACK/NACK is transmitted in the downlink

• ACK/NACK is transmitted using 3 (for 2ms E-DCH TTI) or 12 (for 10ms E-DCH TTI) consecutive slots

Slot #14

Tslot = 2560 chip

bi,39bi,1 bi,0

Slot #0 Slot #1 Slot #2 Slot #i

1 radio frame, Tf = 10 ms

1 subframe = 2 ms


E-RGCH (E-DCH Relative Grant Channel)


• The E-DCH relative grant (RG) a is mapped on E-RGCH:• UP: +1 (possible only in serving E-DCH RLS) • HOLD: 0 (Note: 0 means DTX, i.e, that HOLD is not transmitted)• DOWN: –1

• An orthogonal signature sequence Css,40,m(i),j is assigned to the E-RGCH by higher layers• One slot long orthogonal signature sequence is different in each slot of a 2ms sub-frame

• Higher layers define the sequence pattern• The orthogonal signature sequence is used to identify the E-RGCH from the other E-RGCHs and E-HICHs

(to different UEs) transmitted on the same SF=128 channelisation code channel (there can be at maximum 40 different E-HICHs/E-RGCHs transmitted on the same SF=128 channelisation code channel )

• RG is transmitted using 3 (for 2ms E-DCH TTI from serving E-DCH cell), 12 (for 10ms E-DCH TTI from serving E-DCH cell) or 15 (from non-serving E-DCH cells) consecutive slots

• in each slot is transmitted a sequence bi,0, bi,1, …, bi,39 : bi,j = a Css,40,m(i),j

• Note, that the E-RGCH is a dedicated channel, defined by a channelisation code and a signature sequence, and each E-HICH and E-RGCH transmitted on the same SF=128 channelisation code

• The channel can be independently power controlled by the Node B (vendor specific)

E-AGCH (E-DCH Absolute Grant Channel)

• The E-DCH absolute grant (AG) is mapped on E-AGCH• content of AG and the number of AG bits are FFS• E-AGCH is a shared channel, where only one AG is

transmitted at a time on one SF=256 channelisation code channel

• CRC attachment: 16 bit CRC, masked with the UE identity (E-DCH Radio Network Identifier, E-RNTI, defined by higher layers), is attached to the AG

• Channel coding: rate 1/3 convolutional coding• Rate matching: the number of channel coding output bits is

matched to the number of E-AGCH bits within 2ms• Physical channel mapping: rate matching output bits are

mapped to allocated E-AGCH• E-AGCH has a 2ms sub-frame structure• Note that some simplifications under discussion in 3GPP!

Channel coding

xag,1, xag,2,..., xag,w

Rate matching

ID specific CRC attachment


y1, y2,..., yw+16

z1, z2,..., z3x(w+24)

E-AGCH

r1, r2,..., r60

Coding for E-AGCH


HSUPA and DCH co-existance• For an existing user, E-DCH users will only show as part

of the interference variations (at BTS receiver)• -> Thus mixing DCH & E-DCH users is not a problem• The load variation caused by DCH users are not under BTS

control (but under slower RNC based method)

• Depending on the allocation, there can be allocated both E-DCH and DCH for the same terminal

• E.g. with AMR speech call active while having packet data connection on-going

• This allows smooth introduction for the network as separate carrier is not needed until single carrier capacity fully utilised


E-DCH vs. DCH• DCH does not use BTS based retransmission handling or BTS based scheduling• Both the variable spreading factor principle as well as fast power control

unchanged from DCH to E-DCH• Both support multicode use, though more practical with E-DCH

Feature

Variable spreading factor

Fast power control

Adaptive modulation

BTS based scheduling

DCH

Yes

Yes

No

No

E-DCH

Yes

Yes

No

Yes

Fast L1 HARQ No Yes119 © NOKIA HSDPA/HSUPA.PPT / 22-03-2005 / HH+AT

HSUPA - Summary• Node B based uplink scheduling and HARQ for improved performance• Adaptive modulation not part of HSUPA as power control maintained• HSUPA is backwards compatible and can be introduced gradually in the

network.


HSUPA Capacity


HSUPA Capacity Gain

Uplink cell throughput [kbps]

0200400600800

10001200140016001800

R99 RNC schedulingBLER=1%

HSUPA round robin HSUPA proportionalfair

kbps

Node-B scheduling + faster retransmissions

Release’99

HSUPA

Advanced scheduling

HSUPA


Coverage of high data-rate

Coverage gain 0.5 – 1.0 dB

UE capabilitybeyond 384 kbps

Uplink data rate gain from 384 kbits

(Rel99) to 1-5 Mbits (Rel6)

Capacity gain 20-

50%

Cell throughput

gain

Latency gain<50 ms

Quality of end user

experience

HSUPAHSUPA

Lower costs in transport

Iub capacity gain

Higher add-onPS Traffic

Savings in BB capacity costs

Saves BTS sites (~10%) and adds PS traffic

Savings in transport – in Dedicated VCCsolution max 25%

Higher add-onPS traffic

HSUPA Performance Gains


HSUPA – 3GPP Open Issues


Most of these for WG1, WG2 and WG3 should be closed for the June 2005 spec versions.

For the ASN.1 backwards compatibility not likely to start beforeend 2005 (est.)

3GPP open issues, WG1 (physical layer)


• Compressed mode with 10 ms TTI• Need for new UTRAN/UE measurements or modifications to the existing ones.• Possibility to do slot-by-slot power scaling of E-DPDCH only in case the UE runs

out of transmission power within a TTI.• The exact contents of E-AGCH are not yet defined

• 5 bits for DPCCH to E-DPDCH power ratio is agreed• Total # of bits open, inclusion of time duration, per process flag, etc.• The exact number of bits may also have an impact to the E-AGCH coding and spreading

factor

• E-DPxCH gain factor quantisation.

3GPP open issues, WG2 (MAC, RRC, Stage 2)


• Scheduling behaviour simplifications being discussed• Two schemes are well defined in general but there are some details that still need some

work. • Trying to merge the two schemes as only one

• Details of E-TFC selection and restriction• Priority based scheduling:

• In addition to UE and HARQ based scheduling, Ericsson and NEC are now proposing to have some kind of priority based scheduling where the grants would be applicable up to a given priority only.

• Content of the uplink MAC-signalling scheduling information: • How many bits needed for the buffer occupancy report.

• Transport block sizes: • Agreed to have 4 tables, but now we need to agree on the tables.

• Serving cell change: • L2 based cell selection for faster cell change Nortel and Panasonic propose to have UE

based cell change. Benefits questionable

• Minimum bit rate for 2ms TTI: • One PDU gives already quite a high minimum bit rate (320-bit PDU: 160 kbps)• Potentially too high and something may have to be done to lower it.

3GPP open issues, WG3 (NBAP, FP)


• HARQ Failure indication signalling principles were agreed. • No CR yet implementing the agreement.

• Iub congestion detection was agreed in principle. • No agreement on the exact nature of loss detection. Whether we will use Frame

Sequence Number or Octet sequence number is FFS.

• The big open issue in Congestion notification is the NodeB action to it.

• Agreed to signal Node B a three level congestion notification; no/mild/severe congestion. Will be conveyed in its dedicated control frame.

• Support of 2ms TTI bundling is already there in the frame structure.• Bundling: Bundle five 2 ms AIF packets to one 10 ms FP packet• However, the support is still missing from NBAP and RNSAP.• FFS how / by whom to decide the bundling in case of SHO. Can different branches have

different approach (bundling, no bundling) <-> RNSAP, NBAP

• FP Frame structure, a lot of things done• The needed header CRC length is FFS. Header is getting longer and more CRC bits may

be needed.

3GPP open issues, WG4 (Performance)


• All the performance requirements are still open• Beta factors for requirements and testing on E-DPDCH and E-DPCCH• RAN4 assumptions on FRC, RV index sequence, Switching between different TTI, and

Number of HARQ processes needs to be revisited after RAN1/RAN2 decision• Working points per FRC• Requirements and testcases for E-RGCH and E-AGCH• False alarm rate for E-RGCH• False alarm rate and miss-detection for E-DPCCH• E-HICH, False alarm ACK rate for serving and non-serving cell • Requirements and tests for E-RGCH/E-AGCH• PAR/CM • E-TFC• Active Set Size• Implementation Margin• Finalize ideal simulation results for all channels (DL & UL)• How accurate can the scheduler follow the targets set by the RNC in x% of the time

UTRAN Long Term Evolution


General Requirements for UTRAN Evolution


• Feasibility study started in 3GPP for UTRAN Long Term Evolution with the following requirements

• Packet switched domain optimized• Server UE round trip time below 30 ms and access delay below 300 ms• Peak rates uplink/downlink 50/100 Mbps• Ensure good level of mobility and security• Improve terminal power efficiency • Frequency allocation flexibility with 1.25/2.5, 5, 10, 15 and 20 MHz

allocations, possibility to deploy adjacent to WCDMA• WCDMA evolution work on-going to continue with full speed

• Technologies under consideration include also new radio access technologies as well as new network architecture possibilities

UTRAN Evolution

3.5G HSDPA/HSUPA3GPP Rel.6

3.5G HSDPA/HSUPA3GPP Rel.6


• Best CS + PS combined radio• Current RAN architecture• Spectrum shared with current 3G• Reduced latency• Full service continuity with 2G and

3G

HSPA evolutionHSPA evolution 3.9G new radio access3.9G new radio access

• Optimized for PS only• New architecture• New modulation• Spectrum and bandwidth flexibility• Further reduced latency • Interworking with 3.5G evolution

Documents

Hsdpa Basic